Home / Defined terms / human oversight
3.1 · Corehuman oversight
The human control and monitoring of an AI system across its life cycle, so that a person can understand what the system is doing, judge its output, and intervene or stop it.
What it means here
The human control and monitoring of an AI system across its life cycle, so that a person can understand what the system is doing, judge its output, and intervene or stop it.
The FDIS title uses the term. The public abstract defines it in one sentence: human control and monitoring of AI systems.
How to use it
Use it for the activity and the control, not for the job title. The person who performs it is an overseer.
The term appears in the register wherever the distinction changes what a person must do. A policy that uses the word without the distinction is a policy that cannot be tested. The unit pages link back to this definition so the reader can check the meaning before signing a row.
The units that use it
Every unit in the register is written to be testable, which means the terms in it have to be concrete. Start with the register, then read the unit that carries the term in its requirement. If the term is still unclear after that, the row is not ready to sign.